From street-level delivery to national-scale inclusion
In Mumbai’s relentless traffic, delivery driver Vineet Sawant has one less thing to worry about these days: language barriers. The BBC reported that Sawant, a native Marathi speaker with limited English reading skills, once had to rely on colleagues to interpret delivery instructions written in English. For a job where speed is everything, guesswork was costly.
According to the BBC article, that changed when Sawant’s employer, Zepto — India’s ultra-fast grocery delivery service — partnered with Reverie Language Technologies to integrate artificial intelligence (AI)-powered translation into its driver app. Now, Sawant can switch between six languages, including Marathi, and receive instructions like “ring bell” in words he understands. His daily deliveries have tripled, from 10 to about 30. “When the app speaks our language, we feel more confident, and we work better,” he told the BBC.
Building India-specific AI
India’s linguistic diversity is vast: 22 official languages and hundreds of dialects. Without technology that “understands and speaks” these languages, millions remain excluded from the digital revolution, Professor Pushpak Bhattacharyya of IIT Mumbai told the BBC. The challenge lies in building AI models for languages that lack digitized, high-quality datasets — a stark contrast to English or Hindi, where abundant data exists.
To address this, the Indian government launched Bhashini in 2022 — an initiative developing datasets, AI models, and translation services tailored to the nation’s linguistic diversity. Today, Bhashini hosts 350 AI-based language models, has processed over a billion tasks, and supports 22 languages. It powers multilingual chatbots for public services and translates government schemes into local vernaculars. “Bhashini ensures India’s linguistic and cultural representation by building India-specific AI models rather than relying on global platforms,” says Amitabh Nag, CEO of Digital India’s Bhashini Division.
The goal is bold: Within a few years, rural users could access government services, financial tools, and information systems in their native tongues through voice-enabled AI.
Researchers are also exploring AI for complex, language-sensitive services. At IIT Mumbai’s Koita Centre for Digital Health, Professor Kshitij Jadhav is developing an AI to assist people in quitting smoking — identifying the kind of support each person needs and delivering it in multiple languages, with empathy and emotional nuance. Initial trials are in English and Hindi, but the vision spans all 22 languages. “It will be very customized, not something off the shelf,” Jadhav explained to the BBC.
The human impact
For Sawant, the benefits are immediate and personal. “Not everyone understands English,” he says. “When the app speaks our language, we feel like we belong.” From empowering delivery workers to expanding inclusive public services, India’s AI language initiatives are not just about efficiency — they are about inclusion, dignity, and cultural preservation.
